修复招聘同步会话过期并自动续期

.github/workflows/qq-jobs-sync.yml

@@ -2,7 +2,7 @@ name: 同步招聘信息
 
 on:
   schedule:
-    - cron: "0 */3 * * *"
+    - cron: "17 * * * *"
   workflow_dispatch:
     inputs:
       probe_only:
@@ -49,20 +49,59 @@ jobs:
         with:
           python-version: "3.12"
 
-      - name: 还原专用账号会话
+      - name: 还原滚动专用账号会话
         shell: bash
         env:
+          GH_TOKEN: ${{ github.token }}
+          GH_REPO: ${{ github.repository }}
           QQ_STORAGE_STATE_B64: ${{ secrets.QQ_DOCS_STORAGE_STATE_B64 }}
         run: |
           if [[ -z "$QQ_STORAGE_STATE_B64" ]]; then
             echo "缺少 QQ_DOCS_STORAGE_STATE_B64。" >&2
             exit 1
           fi
           STORAGE_STATE="$RUNNER_TEMP/qq-docs-storage-state.json"
-          printf '%s' "$QQ_STORAGE_STATE_B64" | base64 --decode > "$STORAGE_STATE"
+          SESSION_ARTIFACT_DIR="$RUNNER_TEMP/qq-docs-session-artifacts"
+          mkdir -p "$SESSION_ARTIFACT_DIR"
+
+          session_key="$(printf '%s' "$QQ_STORAGE_STATE_B64" | sha256sum | cut -d ' ' -f 1)"
+          restored=false
+          while IFS= read -r run_id; do
+            candidate_dir="$SESSION_ARTIFACT_DIR/$run_id"
+            if gh run download "$run_id" \
+              --name qq-docs-storage-state \
+              --dir "$candidate_dir" >/dev/null 2>&1; then
+              encrypted_state="$candidate_dir/qq-docs-storage-state.json.enc"
+              if [[ ! -f "$encrypted_state" ]]; then
+                continue
+              fi
+              if SESSION_KEY="$session_key" openssl enc -d -aes-256-cbc -pbkdf2 \
+                -in "$encrypted_state" \
+                -out "$STORAGE_STATE" \
+                -pass env:SESSION_KEY; then
+                restored=true
+                echo "已恢复最近一次成功运行刷新后的专用账号会话。"
+                break
+              fi
+              echo "::warning::滚动会话无法用当前 Secret 解密，继续查找或回退到初始会话。"
+            fi
+          done < <(
+            gh run list \
+              --workflow qq-jobs-sync.yml \
+              --branch "$DEFAULT_BRANCH" \
+              --status success \
+              --limit 20 \
+              --json databaseId \
+              --jq '.[].databaseId'
+          )
+
+          if [[ "$restored" != "true" ]]; then
+            printf '%s' "$QQ_STORAGE_STATE_B64" | base64 --decode > "$STORAGE_STATE"
+            echo "未找到滚动会话，已使用初始专用账号会话。"
+          fi
           chmod 600 "$STORAGE_STATE"
           STORAGE_STATE="$STORAGE_STATE" \
-            python3 -c 'import json, os; json.load(open(os.environ["STORAGE_STATE"], encoding="utf-8"))'
+            python3 -c 'import json, os; state = json.load(open(os.environ["STORAGE_STATE"], encoding="utf-8")); assert isinstance(state.get("cookies"), list) and isinstance(state.get("origins"), list)'
           echo "STORAGE_STATE=$STORAGE_STATE" >> "$GITHUB_ENV"
 
       - name: 安装 Playwright Chromium
@@ -90,6 +129,26 @@ jobs:
           fi
           python3 scripts/sync_qq_jobs.py "${args[@]}"
 
+      - name: 加密刷新后的专用账号会话
+        shell: bash
+        env:
+          QQ_STORAGE_STATE_B64: ${{ secrets.QQ_DOCS_STORAGE_STATE_B64 }}
+        run: |
+          session_key="$(printf '%s' "$QQ_STORAGE_STATE_B64" | sha256sum | cut -d ' ' -f 1)"
+          SESSION_KEY="$session_key" openssl enc -aes-256-cbc -pbkdf2 -salt \
+            -in "$STORAGE_STATE" \
+            -out "$RUNNER_TEMP/qq-docs-storage-state.json.enc" \
+            -pass env:SESSION_KEY
+          chmod 600 "$RUNNER_TEMP/qq-docs-storage-state.json.enc"
+
+      - name: 保存滚动专用账号会话
+        uses: actions/upload-artifact@v4
+        with:
+          name: qq-docs-storage-state
+          path: ${{ runner.temp }}/qq-docs-storage-state.json.enc
+          if-no-files-found: error
+          retention-days: 30
+
       - name: 检查生成文件范围
         id: changes
         if: ${{ !(github.event_name == 'workflow_dispatch' && inputs.probe_only) }}

scripts/sync_qq_jobs.py

@@ -545,6 +545,7 @@ def schedule_capture(response):
                         source_url=_url_with_tab(self.source_url, sheet_id),
                         document_id=DOCUMENT_ID,
                     )
+                await context.storage_state(path=str(self.storage_state))
                 return snapshots
             finally:
                 await browser.close()

tests/test_sync_qq_jobs.py

@@ -1,13 +1,16 @@
+import asyncio
 import base64
 import http.server
 import json
 import sys
 import tempfile
 import threading
+import types
 import unittest
 import zlib
 from html.parser import HTMLParser
 from pathlib import Path
+from unittest.mock import patch
 
 
 ROOT = Path(__file__).resolve().parents[1]
@@ -16,6 +19,7 @@
 from sync_qq_jobs import (
     DatasetSpec,
     IntegrityError,
+    QQDocsSource,
     _resolve_sheet_id,
     assert_matching_scans,
     check_link_accessibility,
@@ -203,6 +207,112 @@ def test_stable_sheet_id_disambiguates_duplicate_view_names(self):
         )
 
 
+class QQDocsSourceSessionTests(unittest.TestCase):
+    def test_complete_collection_persists_refreshed_browser_session(self):
+        class FakePage:
+            def on(self, _event, _callback):
+                pass
+
+            def set_default_timeout(self, _timeout):
+                pass
+
+            async def goto(self, _url, **_kwargs):
+                pass
+
+        class FakeContext:
+            def __init__(self):
+                self.saved_path = None
+
+            async def new_page(self):
+                return FakePage()
+
+            async def storage_state(self, *, path):
+                self.saved_path = Path(path)
+                self.saved_path.write_text(
+                    json.dumps({"cookies": [{"name": "refreshed"}], "origins": []}),
+                    encoding="utf-8",
+                )
+
+        class FakeBrowser:
+            def __init__(self, context):
+                self.context = context
+
+            async def new_context(self, **_kwargs):
+                return self.context
+
+            async def close(self):
+                pass
+
+        class FakeChromium:
+            def __init__(self, browser):
+                self.browser = browser
+
+            async def launch(self, **_kwargs):
+                return self.browser
+
+        class FakePlaywrightManager:
+            def __init__(self, chromium):
+                self.playwright = types.SimpleNamespace(chromium=chromium)
+
+            async def __aenter__(self):
+                return self.playwright
+
+            async def __aexit__(self, *_args):
+                pass
+
+        class ImmediateSource(QQDocsSource):
+            async def _wait_for_workbook(self, _page, _captures):
+                return [{"id": "sheet_daily", "name": "每日更新"}]
+
+            async def _wait_for_pages(
+                self, _page, _captures, _sheet_id, encourage_loading=False
+            ):
+                pass
+
+        snapshot = {
+            "source": {"sheet_id": "sheet_daily", "view_name": "每日更新"},
+            "schema": [],
+            "snapshot": {
+                "source_total": 0,
+                "fetched_count": 0,
+                "pagination_complete": True,
+            },
+            "records": [],
+        }
+        with tempfile.TemporaryDirectory() as directory:
+            storage_state = Path(directory) / "storage-state.json"
+            storage_state.write_text('{"cookies": [], "origins": []}', encoding="utf-8")
+            spec = DatasetSpec(
+                "每日更新",
+                Path(directory) / "daily.json",
+                Path(directory) / "每日更新.md",
+                1,
+            )
+            context = FakeContext()
+            manager = FakePlaywrightManager(FakeChromium(FakeBrowser(context)))
+            async_api = types.ModuleType("playwright.async_api")
+            async_api.async_playwright = lambda: manager
+            playwright_module = types.ModuleType("playwright")
+            playwright_module.__path__ = []
+
+            with patch.dict(
+                sys.modules,
+                {"playwright": playwright_module, "playwright.async_api": async_api},
+            ), patch("sync_qq_jobs.parse_sheet_pages", return_value=snapshot):
+                source = ImmediateSource(
+                    "https://docs.qq.com/smartsheet/example",
+                    storage_state,
+                    timeout_seconds=0,
+                )
+                asyncio.run(source.collect((spec,)))
+
+            self.assertEqual(storage_state, context.saved_path)
+            self.assertEqual(
+                "refreshed",
+                json.loads(storage_state.read_text(encoding="utf-8"))["cookies"][0]["name"],
+            )
+
+
 class MergeHistoryTests(unittest.TestCase):
     def test_first_missing_observation_keeps_record_and_marks_pending(self):
         previous = {